7400 Disk Array

[Nancy Stout]

I have a 7400 SAN and extention fully loaded with 73G drives that are split between 3 servers. One of the servers has 10 30GB luns (serviceguard production server), Second server has 10 30GB luns (QAS server), Third server has 20 15GB luns(DEV server). It was originally configured with AutoRaid. My DB wanted everything configured 0+1. We are at the threshold of going over to raid 5. I have been told that RG1 is at raid 5. We lost a disk last month and had some problems because it was converting everything to raid 5 when we tried to replace the disk. We were down with production for 3 days. Now it is my task to decide if we want everthing put into raid 5 so that this does not happen again. What are the pros and cons for keeping it at 0+1 or converting to raid 5? We are running an SAP environment for financial and HR/payroll with more to be added in the near future.

[A. Clay Stephenson]

It's really very simple: You are trading optimizing for storage capacity (RAID 5) vs. optimizing for performance (RAID 0/1). RAID 0/1 requires two disks for every one disks worth of actual data. RAID 5 is more complicated but a typical value would be 7 disks for 5 disks worth of actual data. The actual value varies with the number of disks used and the parity scheme employed. I've seen write performance differences as large as 5-7X in going from RAID 0/1 to RAID 5 but that seldom happens with modern arrays with lots of cache but there will be some impact.

It appears that you may be nearing capacity limits so that RAID 5 (at least partially) might be your only option. The loss of a disk under either scheme should have resulted in no loss of data as long as the proper disk replacement procedures were followed. I suspect your better answer is to separate production from everything else. Of course that means buying more hardware. It's very easy to do an "oops" and clobber production when all three environments are housed within a common array.

[Luk Vandenbussche]

Nancy,
If you have the VA7400 from AutoRaid to RAID0+1 you need to backup and restore al the data on your VA

[Ted Buis]

The big performance problem with RAID 5DP is the write penalty. So how many writes per second are being done by all the servers, at peak activity? While the buffer cache can help smooth out some of the peaks, if there are more writes/sec than the "back end" of the array can handle, then you will experience reduced performance. Performance will depend on the number and speed of the disks in the array. What is the number (30?) and are the 10K or 15K rpm.

[Nancy Stout]

We have decided for the time being to get us over the possible problems that we had last month to create a larger lun on RG2 to push it over to raid5dp. Does anyone have any ideas about how long it would take to convert the data from raid 0+1 to raid5dp. And possible how big I should make the lun?

Ted, the 30 is the size of the luns that were created for two of the servers. The disks are 73G 10K rpm.

[Ted Buis]

How many physical disks in total? I understood that you had a 7400 (15 drives) plus one(?) expansion bay (15 drives), so I was trying to confirm 30 drives. How many active hot spares in each redundancy group? HP had a tool that would calculate the maximum number of I/Os per second. If you looked at just writes, you could see the reduction in performance potential going to RAID 5DP.

[Nancy Stout]

Ted,

Yes I have 30 drives. Commandview says that Hot spare mode is automatic. I show RG1 active spare is 133.507GB and RG2 active spare is 66.753GB. Is this what you are looking for. If not where do I get the information you are asking for?